Part 2 of 12 in the Try It Yourself series — practical AI guidance for parents at blog.parentsguidetoai.ca
Your child brings home an essay. It's well-written. The sentences flow. The ideas are organized. You're pretty sure they didn't write it — or at least, not all of it. But you don't know how to tell. And you're not sure what to say.
To have that conversation well, you need to understand what AI actually does. Not in a technical way. But in the same way you might understand what Google does for people who use it — well enough to explain it to someone else.
This post covers that. It's shorter than you might expect, because the core idea is actually pretty simple.
It's a very sophisticated autocomplete
The parent's guide at parentsguidetoai.ca describes AI this way, and I think it's the best explanation I've seen for a non-technical audience:
"One way to think about AI is like the autocomplete feature on your phone. It looks at many examples and predicts what is likely to come next."
That's the core of it. When you type a question into ChatGPT or Copilot, the system doesn't "think" about your question the way a person would. It predicts — at enormous speed and scale — what words are most likely to follow, based on patterns it has seen in billions of pieces of text.
It's autocomplete that has read most of the internet.
This is why AI can write a convincing paragraph on almost any topic. It has seen a huge number of examples of how paragraphs about that topic are supposed to sound. It's incredibly good at matching patterns. But pattern-matching is not the same as understanding.
One of the most common ways I use AI on a daily basis is to assist with security analysis of the applications and services I support at work. While the tools are flexible and powerful, they are also confidently wrong quite frequently. In security, we call these false positives, which are bad, because the AI has claimed that a vulnerability exists where there isn't one, but even worse are the false negatives - an AI might be used for analysis and decide there are no security flaws, which could lead to a false sense of security. In order to avoid that, the way that I use AI tools is to augment my workflow, and support the analysis I perform, not to replace the work I do.
Why AI sounds confident even when it's wrong
This is the part that matters most for parents.
Because AI is predicting what words should come next, not checking facts, it can be wrong with complete confidence. It doesn't know it's wrong. It has no way to know. It's producing the most statistically likely response, not the most accurate one.
This has a name in the AI world: hallucination.1 It's when an AI system generates information that is incorrect, made up, or unsupported — but states it as though it were fact.
Hallucinations happen most often with:
- Citations and sources — AI will invent plausible-sounding academic papers that don't exist
- Statistics and numbers — the specific figure sounds right but may be fabricated
- Names and dates — details about real people, events, or places can be quietly wrong
- Local or recent information — things that happened recently, or information specific to a region like BC, are more likely to be wrong2
The danger is that hallucinated information looks identical to correct information. There's no red flag. The AI doesn't say "I'm not sure about this." It just says it.
We'll spend a whole post on how to catch these mistakes (Post 5). For now, the important thing is to know they exist and happen regularly.
What this means for your child's work
Once you understand that AI is a sophisticated pattern-matcher — not a thinker — a few things become clearer.
First, AI-generated writing tends to sound like a lot of other AI-generated writing. Because it's drawing on the most common patterns it has seen, the output tends toward the average. It's competent. It's rarely surprising. Experienced teachers are getting better at recognizing this.
Second, AI can be wrong on factual details — including details in a homework assignment. If your child uses AI to research something and doesn't verify the information, they may submit something that sounds authoritative but contains errors.
Third, there's a real distinction between using AI to understand something and using AI to produce something. The parent's guide puts it well in Section 8:
"A useful question for parents to ask is: 'Could you explain this work without using the AI?'"
That's not a trick question. It's a genuine check. If your child can explain the ideas in their own words, they've learned something. If they can't, the AI may have done the learning for them.
A quick note on what AI is not
The parent's guide is clear on this: "AI systems do not think or understand the world the way humans do."
This is worth saying plainly, because a lot of media coverage treats AI as though it's either a genius or a monster. It's neither. It's a very powerful tool that is very good at certain things and genuinely unreliable at others.
It doesn't have opinions. It doesn't have feelings. It doesn't have goals. When it seems to, that's the pattern-matching at work — it has seen a lot of human text where people express opinions and feelings, so it produces text that sounds like that.
Understanding this doesn't make AI less useful. It makes it more useful, because you know what it's actually good for.
The short version:
- AI works like a very sophisticated autocomplete — it predicts likely words, it doesn't "think."
- This is why it can sound confident and still be completely wrong.
- These errors (called hallucinations) are most common in citations, statistics, dates, and local or recent information.
- The key question to ask your child: "Could you explain this work without using the AI?"
- AI is a powerful tool. It's not a genius, and it's not a threat — it's a pattern-matcher with real limitations.
Now that you have a working mental model of what AI is, the next post gets more practical: we'll walk through the five main tools mentioned in the parent's guide, compare them in plain language, and help you decide which one is worth trying first.
Try it today: Open any AI tool from Post 1 and ask it about a topic you already know well — something from your job, your hobby, or your city. Read the response and notice one thing it gets slightly wrong or oversimplifies. That's the pattern-matcher at work.
Next up: Post 3 — ChatGPT, Copilot, Gemini, Claude, and Perplexity: A Plain-Language Comparison for Parents